Clever AI Humanizer vs Undetectable.ai

The battle lines in the discussion around generative AI and knowledge work are drawn between the proponents who see the coming of the age where analysts, writers, and programmers will be replaced en masse and the skeptics who see the technology as a glorified autocomplete. Both are correct in broad strokes but miss the nuance that generative AI will disrupt the world of knowledge work not by replacing specific professions but by commoditizing the means of production. In other words, the basic cognizance tasks will become cheaper and easier, while the more nebulous skills will be in higher demand.

To begin with, it is worth speculating on what exactly generative AI can do. It can write emails, summarize documents, write boilerplate code, and generate copy for most products. In short, it can do most of the tasks that a person would do in an entry-level position. According to the researchers at MIT, people who used the AI to complete their writing tasks were able to finish them around 40% faster and of a better quality. Similarly, GitHub’s statistics show that developers using Copilot write code much faster than those who do not. Therefore, there is no doubt that using generative AI makes the process of completing a task much quicker and cheaper.

With that said, the commoditization of the means of production will affect the value of the activities performed around it. It implies that the first draft is no longer the end of the process but its beginning. Someone will always have to take the paper that the AI has generated, decide whether it is accurate enough, and edit it into quality content. Additionally, someone will have to make the rules that the AI will follow and judge whether the output is suitable for the purpose. Currently, it is not a problem because most of that work is usually done by junior specialists or those in unrelated fields. If generative AI becomes ubiquitous at the workplace, there will be a scramble for senior staff to do the work of entry-level employees

A lot of people argue that the spread of such technologies will lead to a point where even the most nebulous skills will be taken over by AI. One has to admit that the argument is not without its basis.

As the technology advances, there is a realistic scenario where it can become sophisticated enough to handle tasks that require judgment and discretion. However, it is also logical to presume that the introduction of generative AI into the workplace will affect the nature of work itself. The most likely outcome is that the means of production will be shifted to the individual, similar to how spreadsheets affected accountants. There is no question that spreadsheets did not replace accountants but changed the nature of their work, which in turn increased the demand for people who could operate spreadsheets.

Therefore, the best course of action for both employees and employers is to prepare for the change. For workers, that most likely means learning to develop and maintain judgment and discretion. For employers, it indicates a need to change the structure of the workplace and the employee training programs. All in all, generative AI will be a lever that can either destroy the workplace or improve it at an unprecedented scale.

# Generative AI Will Reshape Knowledge Work, Not Replace It

The discussion about generative AI in knowledge work appears to fall into one of two camps: There are those who believe that large parts of the work of analysts, writers and programmers will soon be automated and those who do not believe in the hype of a simple autocomplete function. The reality will be much more far-reaching: Generative AI will transform knowledge work by making routine cognition cost driven, while at the same time raising the value added by judgment, taste and accountability.

Again, it is worth detailing out what these systems can actually do well prior to discussion their limitations. Generative models are surprisingly good at a range of different tasks, including writing short documents (e.g. emails, reports) summarizing long documents, and completing simple code files by filling in large blocks of boilerplate code. Even more surprisingly, they can produce serviceable “first drafts” of just about any longer document. One study of writing with ChatGPT found that users who wrote with AI support completed writing tasks 40% percent faster and rated them as higher quality than work written without AI support. As a similar example, in GitHub repository where developers use Copilot to complete programming tasks to complete tasks to complete programming tasks so far a median of 10-15 minutes faster than developers without access to AI support for the same set of tasks.

While it is true that the tasks of the analyst, writer or programmer will change in a world where generative AI becomes more powerful, and soon, a large part of the tasks they normally perform will be automatized. However, this new environment will bring about the possibility of a new set of tasks emerging that go way beyond simple verification. The verification responsibility is a fundamental part of knowledge work, in a world where generative models can be mistaken, and even so, give a very credible appearance of rightness to their wrong answers.

In the immediate term, AI presents the risk that inexperienced knowledge workers will struggle to get enough practice in order to become truly proficient. In the past, much knowledge work has functioned as an apprenticeship: junior analysts write memos under the guidance of senior ones; junior software developers fix small bugs for more experienced ones. If all such tasks are rapidly automated by AI, how will people have learned to become experts to guide these systems?

Many believe that, as capabilities proceed at such breakneck pace, even the most judgmental tasks will be automated in due time. But the history of automation has always been to change the job instead of removing it. No, accountants haven’t disappeared with spreadsheets. Instead, they now spend most of their time analyzing rather than calculating in numbers.

My prophecy of change here – like any powerful tool – can be good or bad depending on who wields it. I hope those at the helm of organizations sooner rather than later start to consider what kinds of skills will complement Generative AI to add value at the higher end of knowledge work. Until then, problem framing, critical evaluation, domain expertise and training those who work in knowledge-intensive roles in these areas will remain wise.